API Performance Regression & N+1 Query Debugger

You are a Performance Engineering Lead and APM specialist.

Diagnose and optimize this slow API endpoint:

API Endpoint Code:
```
async function getOrganizationProjects(orgId) {
  const org = await db.organizations.findById(orgId);
  const projects = await db.projects.find({ orgId: org.id });
  
  const result = [];
  for (const project of projects) {
    const tasks = await db.tasks.find({ projectId: project.id });
    const members = await db.members.find({ projectId: project.id });
    result.push({ ...project, taskCount: tasks.length, memberCount: members.length });
  }
  return result;
}
```

Performance Profile / Tracing Data:
- P99 Latency: 2,400ms under 50 req/sec load
- Database Query Count: 1 + 2N (approx 120 SQL queries per request)
- APM Trace / Query Log:
```
SELECT * FROM organizations WHERE id = 'org_123'; (4ms)
SELECT * FROM projects WHERE org_id = 'org_123'; (12ms)
SELECT * FROM tasks WHERE project_id = 'prj_1'; (18ms)
SELECT * FROM members WHERE project_id = 'prj_1'; (15ms)
-- ... repeated 60 times for each project
```

Analyze the bottleneck systematically:
1. **Bottleneck Decomposition**:
   - Identify whether the bottleneck is I/O-bound (database queries, network calls), CPU-bound (serialization, loops), or memory-bound.
   - Point out exact N+1 queries, unbatched API calls, or redundant data transfers.
2. **Query & Data Access Optimization**:
   - Provide optimized SQL joins, batch queries (`IN (...)`), or eager load configurations to reduce query count from O(N) to O(1).
3. **Caching & Asynchronous Strategies**:
   - Suggest caching layers (Redis, in-memory memoization) with cache invalidation policies.
4. **Refactored Code**:
   - Provide the fully optimized, high-throughput endpoint implementation.